AI-Mediated Mental Health Support: The Role of Attachment Orientation and Psychological Distress

Key Findings

• — 29.3% of respondents (171 of 584) reported turning to general-purpose generative AI at least once for emotional support, coping advice or mental health assistance; the proportion was 47.0% among those reporting severe distress
• — Psychological distress was the strongest psychological correlate of use; younger age and female gender also predicted use
• — Attachment anxiety was associated with use, but the association was no longer significant once distress was entered, which the authors attribute to variance shared between the two constructs
• — Attachment avoidance was negatively but not significantly associated with use
• — The authors state that the high uptake among severely distressed users points to a need to evaluate the safety, crisis-response quality and clinical-displacement risks of consumer-facing AI

Methodology Notes

Nonprobability online panel sample (iPanel, Israel), data collected November 2025; adults 18-70 who had used a general-purpose generative AI tool such as ChatGPT or Gemini in the past year; 16 respondents excluded on attention checks; final n=584 (mean age 43.5, 49.5% female). Use was a single self-reported binary item that counts anything from one exchange to sustained use. Published online 2026-09-29 (Crossref; PubMed PMID 42808327, entrez date 2026-09-29). The SAGE page returned 403, so the abstract was read from Crossref and PubMed, and methods and odds ratios from the PsyArXiv preprint (posted 2026-02-18). Preregistration is stated in the preprint. The preprint's final model reported avoidance as marginally significant (p=.050); the published abstract reports it as non-significant, and the key findings follow the published abstract.

Sources